- aktuell - lists.lrz.de

Upd:Linux-Cluster: All partitions available again
by LRZ aktuell May 3, 2012

May 3, 2012

Changes to this message: Update Update: The cluster and visualization systems are available again. In the Myrinet segment, quite a number of systems are still offline, but batch processing should work again. The login nodes lx64ia2 and lx64ia3 are available for submission of SGE jobs; however Parastation MPI will not run on lx64ia2 any more. Please login to the MPP cluster and build and run your programs from there as indicated in the cluster docs. ----------------------------------------------------------------------- Dear users of the Cluster systems at LRZ, as announced in http://www.lrz.de/aktuell/ali4261/ user operation on the cluster will be interrupted due to an electrical power shutdown test which affects all storage systems. In order to cleanly shut down all accesses to the storage systems, we have already started to drain the SLURM queues and all running jobs will be removed from 4:30 pm onwards. Please also note that after this interruption, the SGE queues mpi_8, shm_8, mpi_32 and shm_32 will become permanently unavailable because the systems will be upgraded and moved to SLURM. The SLURM documentation will be appropriately updated as soon as this upgrade is completed. We will keep you up-to-date about the status of the cluster via this document. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/aktuell/ali4295/ Reinhold Bader

1 0

Upd:Linux-Cluster: All partitions available again
by LRZ aktuell May 3, 2012

May 3, 2012

Changes to this message: Update Update: The cluster and visualization systems are available again. In the Myrinet segment, quite a number of systems are still offline, but batch processing should work again. The login nodes lx64ia2 and lx64ia3 are available for submission of SGE jobs; however Parastation MPI will not run on lx64ia2 any more. Please login to the MPP cluster and build and run your programs from there as indicated in the cluster docs. ----------------------------------------------------------------------- Dear users of the Cluster systems at LRZ, as announced in http://www.lrz.de/aktuell/ali4261/ user operation on the cluster will be interrupted due to an electrical power shutdown test which affects all storage systems. In order to cleanly shut down all accesses to the storage systems, we have already started to drain the SLURM queues and all running jobs will be removed from 4:30 pm onwards. Please also note that after this interruption, the SGE queues mpi_8, shm_8, mpi_32 and shm_32 will become permanently unavailable because the systems will be upgraded and moved to SLURM. The SLURM documentation will be appropriately updated as soon as this upgrade is completed. We will keep you up-to-date about the status of the cluster via this document. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/aktuell/ali4295/ Reinhold Bader

1 0

WLAN-Verbesserung im Maschinenwesen
by LRZ aktuell May 2, 2012

May 2, 2012

Durch die stark zunehmende Zahl von Studierenden sowie der mobilen Geraete war die WLAN-Versorgung im Gebaeude des Maschinenwesen in Garching nicht mehr ausreichend. Die Studienbeitragskommission der Fakultaet Maschinenwesen hat deshalb knapp 70.000 Euro fuer die Verbesserung bereitgestellt. Gemeinsam mit dem LRZ wurde ein Konzept erarbeitet und teilweise bereits umgesetzt. In einem ersten Schritt wurden in allen oeffentlichen Bereichen die bestehenden APs (Accesspoints) durch neueste Modelle mit einer Bandbreite von bis zu 400 Mbit/s ersetzt. In den Hoersaelen, in denen die Anzahl gleichzeitiger Nutzer naturgemaess noch hoeher ist, wurden zusaetzliche APs zur Verstaerkung installiert. Insgesamt wurden ueber 70 APs verbaut. Durch die im letzten Jahr extrem zunehmende Zahl an Nutzern wurden auch die fuer WLAN verwendeten IP-Netze zu klein. Da die bisherigen Switches in den Gebaeudehauptverteilern nur 30 verschiedene Netze (VLANs) unterstuetzten und die Lehrstuehle, Rechnerpools, etc. jeweils auch eigene Netze benoetigen reichte dies nicht mehr aus. Als Konsequenz mussten in den acht Gebaeudehauptverteilern die Zentralswitches erneuert werden. Da aus Brandschutzgruenden keine zusaetzlichen APs in der Magistrale installiert werden duerfen, konnte die Planung des LRZ in diesem Bereich nicht umgesetzt werden und die Versorgung in der Magistrale ist deshalb noch nicht optimal. Es ist jedoch eine Loesung erarbeitet worden, welche nur passive Antennen im Fluchtwegbereich vorsieht. Wegen aufwaendiger Leitungsinstallationen ist die Umsetzung aber voraussichtlich erst bis zum Beginn des Wintersemesters 2012/2013 moeglich. Der Zeitplan fuer die erledigten Arbeiten war sehr ambitioniert. Die Mittel wurden Mitte Februar bewilligt und die Ersetzung konnte komplett in der vorlesungsfreien Zeit bis zum 15. April abgewickelt werden. Dies war auch nur moeglich, weil im LRZ neben den verantwortlichen Mitarbeitern drei Auszubildende zusaetzlich mit dem Loewenanteil ihrer Arbeitszeit fuer das Projekt gearbeitet haben. Diese Information finden Sie im WWW unter http://www.lrz-muenchen.de/services/netz/aktuell/ali4299/ Wolfgang Beyer

1 0

PRACE Summer School on Code Optimisation for Multi-Core and Intel(R) MIC Architecture
by LRZ aktuell May 2, 2012

May 2, 2012

The Partnership for Advanced Computing in Europe (PRACE) is pleased to announce the Summer School on Code Optimisation for Multi-Core and Intel Many Integrated Core (MIC) architecture at the Swiss National Supercomputing Centre in Lugano, Switzerland. In this three day intensive event participants will focus on programming and tuning techniques for modern multi- and many-core processors with a particular focus on the Intel Many Integrated Core architecture. Topics will be presented at an advanced level and will include structuring code to enable SIMD/vectorization, efficient usage of register, cache and memory hierarchies, use of multi-threading techniques to maximize resource utilization, data locality considerations on multi-socket NUMA nodes, and inter-node communication. Intel specialists will introduce the Intel MIC architecture and the Intel programming environment, and will delve into greater detail of the use of multi- and many-core programming techniques. Demonstrations and hands-on sessions will be integrated throughout the course to illustrate the topics in greater depth, and user case studies will highlight real-world experiences in optimising scientific (HPC) codes on the Intel MIC architecture. Dates ===== 21st-23rd June 2012 Location ======= Swiss National Supercomputing Centre, Lugano, Switzerland Registration ========== To apply for a place please go to http://survey.ipb.ac.rs/index.php?sid =43493&=en. The school is offered free of charge to students and academics residing in PRACE member states, and lunches, coffee breaks and an evening dinner event will be provided. Please note that a sound knowledge of C, C++ or Fortran is a pre-requisite for this course. Attendance is limited to 30 places, and selection of participants will be based on demonstrated understanding of computational science and high-performance computing and experience with at least one of MPI, OpenMP, or multi-core programming. Registration closes Friday 11th May 2012, but we encourage applicants to apply as early as possible. Candidates will be informed of the results of their application by the second week of May (at the latest). Registration deadline: 11 May 2012. Please contact tim.robinson(at)cscs.ch for further information or see http://prace-portal.cscs.ch/blog/?tx_typopress_pi1[post]=138 This information is also available on our web server http://www.lrz-muenchen.de/services/compute/supermuc/aktuell/ali4298/ Alexander Block

1 0

Upd:Linux-Cluster: Mostly returned to user operation
by LRZ aktuell April 27, 2012

April 27, 2012

Changes to this message: UV3 available again Update: The cluster and visualization systems are mostly available again, with the following exception: * the Myrinet cluster is under maintenance as indicated below and will not be available again before early next week. The login nodes lx64ia2 and lx64ia3 are available for submission of SGE jobs; however Parastation MPI will not run on lx64ia2 any more. Please login to the MPP cluster and build and run your programs from there as indicated in the cluster docs. ----------------------------------------------------------------------- Dear users of the Cluster systems at LRZ, as announced in http://www.lrz.de/aktuell/ali4261/ user operation on the cluster will be interrupted due to an electrical power shutdown test which affects all storage systems. In order to cleanly shut down all accesses to the storage systems, we have already started to drain the SLURM queues and all running jobs will be removed from 4:30 pm onwards. Please also note that after this interruption, the SGE queues mpi_8, shm_8, mpi_32 and shm_32 will become permanently unavailable because the systems will be upgraded and moved to SLURM. The SLURM documentation will be appropriately updated as soon as this upgrade is completed. We will keep you up-to-date about the status of the cluster via this document. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/aktuell/ali4295/ Reinhold Bader

1 0

PRACE Preparatory Access Call for Tier-0
by LRZ aktuell April 27, 2012

April 27, 2012

Preparatory access is intended for testing and developing codes in order to prepare applications for PRACE Tier-0 project access. In particular, we would like to draw your attention to the Preparatory Access Type C which includes code development with support from PRACE experts. Further information is given on the PRACE web page http://www.prace-ri.eu/Call-Announcements The Preparatory Access Call is a rolling call, researchers can apply for resources all year. There are no closing dates, proposals will be reviewed every third month. The next cutoff date is June 1 2012. LRZ invites all interested scientists, whose projects fit the criteria of the PRACE calls to submit their projects. Further information about PRACE calls as well as information on the PRACE machines, can be found at http://www.prace-ri.eu/ If you need some extra information please feel free to contact PRACE directly via peer-review(a)prace-ri.eu or weinberg(a)lrz.de This information is also available on our web server http://www.lrz-muenchen.de/services/compute/aktuell/ali4297/ Volker Weinberg

1 0

Upd:SuperMIG back in production (TSM also available now)
by LRZ aktuell April 27, 2012

April 27, 2012

Changes to this message: Update Dear users of SuperMIG, the power cut-off tests and SuperMIG maintenance are completed and the system is back in production. Update: The move of the TSM server into another computer room is also finished and TSM is available on SuperMIG. We apologize for the inconvenience caused by this interrupt. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/supermuc/aktuell/ali4294/ Hans-Georg Kleinhenz

1 0

Upd:SuperMIG back in production (TSM not available)
by LRZ aktuell April 26, 2012

April 26, 2012

Changes to this message: Update Dear users of SuperMIG, the power cut-off tests and SuperMIG maintenance are completed and the system is back in production. TSM is not yet available on SuperMIG, since the TSM server is being moved into another computer room. We expect TSM to become available tomorrow (April 27th) and will give you an update through this ALI. We apologize for the inconvenience caused by this interrupt. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/supermuc/aktuell/ali4294/ Hans-Georg Kleinhenz

1 0

Upd:Linux-Cluster: Mostly returned to user operation
by LRZ aktuell April 26, 2012

April 26, 2012

Changes to this message: Mostly available again Update: The cluster and visualization systems are mostly available again, with the following exceptions * uv3 could not be started and will be repaired as soon as possible * the Myrinet cluster is under maintenance as indicated below and will not be available again before early next week. The login nodes lx64ia2 and lx64ia3 are available for submission of SGE jobs; however Parastation MPI will not run on lx64ia2 any more. Please login to the MPP cluster and build and run your programs from there as indicated in the cluster docs. ----------------------------------------------------------------------- Dear users of the Cluster systems at LRZ, as announced in http://www.lrz.de/aktuell/ali4261/ user operation on the cluster will be interrupted due to an electrical power shutdown test which affects all storage systems. In order to cleanly shut down all accesses to the storage systems, we have already started to drain the SLURM queues and all running jobs will be removed from 4:30 pm onwards. Please also note that after this interruption, the SGE queues mpi_8, shm_8, mpi_32 and shm_32 will become permanently unavailable because the systems will be upgraded and moved to SLURM. The SLURM documentation will be appropriately updated as soon as this upgrade is completed. We will keep you up-to-date about the status of the cluster via this document. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/aktuell/ali4295/ Reinhold Bader

1 0

Upd:Linux-Cluster: Mostly returned to user operation
by LRZ aktuell April 26, 2012

April 26, 2012

Changes to this message: Mostly available again Update: The cluster and visualization systems are mostly available again, with the following exceptions * uv3 could not be started and will be repaired as soon as possible * the Myrinet cluster is under maintenance as indicated above and will not be available again before early next week. The login nodes lx64ia2 and lx64ia3 are available for submission of SGE jobs; however Parastation MPI will not run on lx64ia2 any more. Please login to the MPP cluster and build and run your programs from there as indicated in the cluster docs. ----------------------------------------------------------------------- Dear users of the Cluster systems at LRZ, as announced in http://www.lrz.de/aktuell/ali4261/ user operation on the cluster will be interrupted due to an electrical power shutdown test which affects all storage systems. In order to cleanly shut down all accesses to the storage systems, we have already started to drain the SLURM queues and all running jobs will be removed from 4:30 pm onwards. Please also note that after this interruption, the SGE queues mpi_8, shm_8, mpi_32 and shm_32 will become permanently unavailable because the systems will be upgraded and moved to SLURM. The SLURM documentation will be appropriately updated as soon as this upgrade is completed. We will keep you up-to-date about the status of the cluster via this document. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/aktuell/ali4295/ Reinhold Bader

1 0